| |
Ralph Hempel <rhempel@bmts.com> wrote:
> I just figured out this weekend (thanks Kekoa) that the H8/300 stack
> pointer (r7) and the power-down function in the RCX have a close
> relationship.
>
> The stack pointer MUST be in the on-chip RAM area (0xFD80 to 0xFF7F)
> before calling the RCX shutdown_power() function. This is because the
> shutdown_power() funtion (I think) deselects the external RAM to put
> it into a low-power mode before putting the CPU itself to sleep.
>
> If r7 is pointing outside of the on-chip RAM, then it cannot be
> retrieved and the RCX will hang, and in such a nasty way that you
> will have to remove the batteries for about 5 minutes - not a few
> seconds like normally.
>
> This is especially important for users of legOS and pbFORTH since we
> are likely to allocate task stacks or put the stack pointer outside of the
> small 512 byte stack area.
After talking with Ralph over the weekend, I suggested a test he might do
to figure out if port 5 bit 2 really does what he suggested it might. I
don't think he did this test, so I fired up my RCX and hacked it together.
I verified that port 5 bit 2 does indeed appear to put the RAM in low-power
mode.
The tests were simple. First test: move stack to on-chip RAM, verify that
power_off is broken. Second test: move stack to on-chip RAM, replicate the
ROM power_off function in asm, remove the code that puts RAM into low power
mode, call modified power_off instead of ROM power_off, verify that new
power_off works.
The problem with having RAM in low-power mode when returning from sleep (if
this is indeed what is going on) is subtle. The instruction after the
sleep lives in ROM and turns the RAM back on. At first glance, the RAM
does not seem to be accessed until the rts from power_off, when the return
address is fetched from the stack. This is not the case. Sleep ends on
IRQ1; the interrupt handler is called before power to RAM is restored, and
the stack is needed at this time.
-Kekoa
|
|
Message has 2 Replies:
 | | Re: H8300 Stack and SLEEP
|
| Sorry I'm late. I believe this is not an issue with legOS. legOS only calls the ROM powerdown function at the end of its kmain() loop, where the stack pointer should be in the on-chip RAM area anyway. When sleeping in the idle task, legOS directly (...) (28 years ago, 20-Mar-99, to lugnet.robotics.rcx.legos)
|
Message is in Reply To:
4 Messages in This Thread:
  
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
